In today's episode, we show how to turn AI from a bland, generic appliance into a highly personalised tool that actually speaks and thinks like you by treating models as temporary environments and encoding your identity in local NLP-based “voice files” and an AI brain.

I walk through Heather’s method — granular written voice specs (including lists of forbidden words), logical levels mapped to simple markdown files on your machine, and the idea that ChatGPT or Claude are just swappable ovens in your kitchen.

We give concrete examples: a Language Pattern Practice app that gamifies audio drills (yes, people were doing it at 11pm — witchcraft, I tell you) and a Strategy Detective that runs tote loops to pull out submodalities via a voice-only interface.

Crucially, we stress that AI is a lightning-fast pattern generator but blind to micro‑cues, so the human retains veto power with a simple keep/edit/reject workflow — reject is practically sacred. That flips your professional value from churning content to exercising real-world discernment: you’re the master architect, not the machine.

For a practical nudge, Heather challenges you to name one thing you’d happily hand over to AI and one thing you’d never outsource, and to explain why — that “why” is your irreplaceable X‑factor.

Oh, and meta moment: this episode is AI-generated from the Start With AI Newsletter on LinkedIn using NotebookLM, so we’re practising what we preach with a cheeky grin.

The Details

If you want practical takeaways, we get delightfully tactical. I walk you through how to start building your own AI brain today: pick a handful of core phrases you love, list words you hate, dump them into simple markdown files on your laptop, and point whatever chat tool you’re using at that brain. Example: forbid “delve” and “unprecedented” and watch your outputs stop sounding like corporate stock photos.

Then try Heather’s Language Pattern Practice app idea — make repetition audio‑based and gamified so students actually do their homework (yes, people will practise language drills at 11pm if it’s fun).

For coaches and training programmes, outsource the boring repetition to the AI game and keep the live session for human calibration and reading the room. We also give a short workflow for coaches:

(1) choose one repeatable task you trust the AI with (homework quizzes, pattern drills),

(2) choose one thing you won’t hand over (real‑time empathy, reading micro‑expressions),

(3) codify your voice and values in markdown,

(4) use the AI to generate options and always keep the veto. The Strategy Detective app is a beautiful template: voice‑only prompts keep people inside the memory, elicit submodalities (brightness, movement, sound) and build anchors.

Final nudge from us: do Heather’s two‑question exercise — what you’d hand over and what you’d guard — because that “why” is the secret sauce of your professional brand.

Chapters:

• 00:12 - The Signature of AI-Generated Writing
• 01:38 - Mission: Personalizing AI — Making Models Sound Like You
• 06:21 - Creating an AI Brain: Solving Copy-Paste Fatigue and Platform Lock-In
• 09:46 - From Theory to Practice: Two Concrete NLP Applications
• 14:43 - Where Humans Still Matter: The Practitioner’s Role
• 19:36 - From Theory to Practice: Applying AI to Education & Training

Takeaways:

• We learned that a written voice file — not an audio clip — can shove that bland, corporate stock-photo voice right out and make the AI sound like you, not like a robot.
• I realised that building an 'AI brain' with simple local markdown files keeps our identity portable, prevents vendor lock-in, and means we update once and every model happily reads our rules.
• We explored how mapping Neuro-Linguistic Programming's logical levels onto AI turns environments like ChatGPT into swappable tools, while identity, beliefs and values live in our files so the model follows us, not the other way round.
• I loved the Strategy Detective idea: a voice-only tote loop that guides you through sensory submodalities of success, helping you recreate winning behaviours without the keyboard yanking you out of the memory.
• We agreed the human keep/edit/reject power is priceless — AI can spit out a hundred shiny options, but only we can read micro-expressions and veto what doesn't actually fit the person in front of us.

Companies mentioned in this episode:

• ChatGPT
• Claude
• Grok
• Twitter

You know that feeling when you open an email or maybe a blog post, and within maybe two or three sentences, your brain just kind of flags it?

Like, you immediately know an AI wrote it.

Oh, absolutely.

It definitely has a very specific signature.

It carries this perfectly polite, mathematically flawless, but somehow completely sterile tone.

Right.

I always describe it as a corporate stock photo, but, you know, constructed entirely out of text.

Yeah, that's a perfect way to put it.

It's that I hope this email finds you well in these unprecedented times.

Energy, but stretched across an entire article.

Exactly.

It's technically perfect, but literally nobody actually talks like that in the real world.

And for those of you listening right now, I mean, you've probably just accepted this, right?

We all kind of have.

There's this collective unspoken agreement that this robotic, predictable output is just.

Well, it's just the toll we pay on the AI highway.

Yeah.

Like, if you want the superhuman speed and convenience, you just have to sacrifice the messy, authentic humanity.

You just hand it over.

Which is the default assumption in the tech world today.

I mean, the general public has been entirely conditioned to view AI as this.

Generic appliance, like a toaster or something.

Right.

It's a microwave.

You push a button, you get a standardized result, and you just adapt yourself to how the machine works.

But then you stumble across a piece of reading that completely shatters that assumption.

And that is exactly what we are getting into today.

Welcome to this deep dive, everyone.

Glad to be here.

Our mission today is to figure out how to transform AI from that generic, flavorless appliance into a highly personalized tool that actually sounds, and more importantly, thinks, thinks exactly like you do.

And we are doing this by exploring a really fascinating methodology developed by a practitioner named Heather.

Yes, we are going to be unpacking some excerpts from her document titled, your NLP training manual is already an AI Spec Part one.

The central premise here, and this is the hook that should honestly completely change how you view your relationship with technology, is that instead of treating AI as a finished product, we just use, Heather argues, we need to treat it as a raw material that we design.

Right.

We are the designers.

It is a massive paradigm shift.

I mean, it essentially wrestles the power of personalization away from those developers sitting in some lab in Silicon Valley, and.

It puts it directly into the hands of the individual sitting right at their own laptop.

Exactly.

So what does this design process actually look like in practice.

And how do we scrub that generic, you know, stock photo voice out of the machine?

Well, Heather starts the document with this really foundational demonstration from a course she's building.

It's called the AI Model Yourself Boot Camp.

Right.

She uses a very straightforward A B test to prove her point here.

Yeah, so in the first part of this test, she just feeds an AI model a standard prompt.

Like, she gives it the instructions for a specific writing task, hits generate, and just lets it do its thing.

And as you'd expect, the output is that very safe, unmemorable text we were just complaining about it.

Gets the job done, but it leaves absolutely zero impression on the reader.

Just completely forgettable.

Okay, let's unpack this.

Because the second half of the test is where the magic at actually happens.

Right.

Version B.

For version B, she doesn't just give the AI the prompt.

She pairs it with something she calls a voice file.

And just to be super clear for you listening, this isn't an mp3 recording of her speaking or anything?

No, not at all.

It's a written specification.

It's this highly granular document that models exactly how she thinks, what her core values are, the specific phrases she leans.

On, and crucially, this explicit list of words she would never say.

Yes, giving the system a list of forbidden words is incredibly effective.

I found that so interesting.

It really is, because most people try to prompt AI by telling it what to add, right?

Like, be more funny or be more casual.

But giving it a negative constraint forces the model to abandon its default pathways.

Right.

By telling it under no circumstances use the word delve or unprecedented, you're forcing the algorithm to actually work for its vocabulary.

Exactly.

It's basically like sculpting.

By removing the negative space.

The analogy that comes to my mind here is acting.

Giving an AI a standard prompt is like handing an actor a script and just saying, read these words.

That's a great way to look at it.

Right?

But giving the AI this deeply detailed voice file is like handing that same actor a fully fleshed out character backstory.

You give them their motivations, their childhood fears, their pet peeves, so suddenly they.

Aren't just reciting lines anymore, they are actually embodying a role.

Exactly.

That analogy captures the essence perfectly.

You're providing a Persona for the AI to inhabit while it executes the task.

And the results of this AB test in her document are night and day.

Totally.

In the second version, guided by that voice file, the author notes that the real messy human suddenly shows up in.

The Language, the messy human.

I absolutely love that phrasing.

So instead of a sterile conclusion, the AI might actually, you know, invite the reader to kick back and grab a coffee.

Purely because that's a real idiosyncrasy documented in the user's voice file.

It introduces actual texture.

It breaks the sterile mold because it has been given the actual boundaries of a real person's communication style.

But hold on.

If I'm thinking about the logistics here.

Yeah.

Wait, isn't there a risk of becoming too dependent on one platform if you build this elaborate setup?

Like a voice file is a fantastic trick for one prompt.

But if I'm jumping between chatgpt on my phone and Claude for my writing and maybe Grok for something else, I.

See where you're going with this.

Do I have to copy and paste this massive character backstory every single time I open a new chat window?

That sounds absolutely exhausting.

A single file doesn't really seem like a robust system.

And that raises a really important question.

It's exactly where Heather's approach moves from just a neat prompting trick into actual software architecture.

Okay.

She solves this copy paste fatigue by creating what she calls an AI brain.

And she structures this brain using a foundational NLP concept known as logical levels.

Okay, my brain just did a bit of gymnastics there.

I know it sounds wild.

Break down what logical levels actually means in a human context first before we map it to all this tech.

Sure.

So in nlp, the concept of logical levels helps explain how humans function and change, basically moving from the outside world inward.

Okay.

At the very bottom, base level, you have your environment, like where you are and who is around you.

One level up is behavior, the specific actions you take.

Got it.

Above that are your capabilities, the skills you possess to do those actions.

Higher still are your beliefs and values, like why you do what you do.

And at the very top, dictating all those levels below it is your identity, who you fundamentally believe you are.

Okay, so just to make sure I'm tracking, if I'm trying to become a runner.

The environment is the park trail.

The behavior is jogging.

The capability is knowing how to pace myself.

The belief is that cardio is good for my longevity.

And the identity at the very top is just, I am an athlete.

That is exactly how the psychological model works.

If you change your identity at the top, it naturally filters down and changes your behaviors and your environment.

At the bottom, that makes total sense.

Now, take that human psychological model and lay it over modern artificial Intelligence.

Oh, wow.

Okay, so how does she actually map that?

Well, Heather treats the AI models themselves.

ChatGPT, Claude.

Whichever platform you use merely as the environment, they sit at the very bottom level.

Oh, I see.

They are temporary spaces.

They get updated, they change features, they become obsolete.

But your identity, your voice, your brand, your business rules, that is stored locally on your own computer in the form of incredibly simple markdown files.

Which are just like basic unformatted text files, right?

Precisely.

So your identity sits at a higher logical level than the software you're actually using.

The AI brain is a central repository on your hard drive.

That is fascinating.

When you change a core belief or tweak a service offering, or realize there's a new language pattern you want to use, you update your local text files just once.

Then you simply point whichever temporary environment you happen to be using today at.

That brain, and the platform just reads.

It adapts to your identity and operates within your parameters.

That completely eliminates the anxiety of picking the wrong tool.

Totally.

Because I was genuinely worried that if I spent months training one specific AI and then a competitor releases a far superior model next year, I'd just be trapped.

Like I'd have to start all over.

Exactly.

And that's what this architecture prevents.

Because the platform is just the environment, it becomes entirely swappable.

It keeps the power with the practitioner.

You aren't locking your deepest brand secrets into a proprietary tech ecosystem where they might be used to, you know, train other models or just drift out of date.

The environment swaps out, but the identity remains uniquely yours.

It's brilliant theory, but moving from theory to practice.

Having an organized brain is great, but how does this architecture translate into actual functioning tools?

Does the text give examples?

Oh, yeah.

She provides two very concrete applications she built using these NLP principles.

Let's look at the first one, which she calls the Language Pattern Practice App.

I saw this in the text.

I know it involves a database of language patterns, but what really struck me is that she turned something highly academic into a game.

Gamification is the key here.

And she specifically incorporates an audio element.

So the user actually has to speak the patterns out loud.

Right, because reading it isn't enough.

No.

In NLP training, reading a pattern in a manual doesn't do much good.

You really need repetition.

The pattern has to move from the page into your ear and finally become muscle memory in your daily speech.

So making it an interactive audio based game just completely removes the friction of.

Studying it, actually makes it addictive.

She notes in the document that people from her Practice groups were voluntarily logging into this app to run through language drills at 11 o' clock at night.

Wait, really?

11 O' clock at night?

Yeah.

Getting people to do psychological homework at 11pm voluntarily is practically witchcraft.

That alone is incredibly impressive.

It really is.

But the second application she details is the one that really grabbed my attention.

She calls it the Strategy Detective.

Ah, yes.

The Strategy Detective is a fascinating tool.

It is built entirely around another core NLP framework called the tote.

Right.

How does a tote loop actually operate?

The easiest way to visualize a tote loop is to think of a digital thermostat in your house.

Okay, so the thermostat's goal is to keep the room at 70 degrees.

First, it tests the environment.

Maybe it reads the room temperature at 65 because there's a gap between the goal and reality.

It operates.

It turns on the furnace.

Makes sense.

Then a few minutes later, it tests again.

Is it 70 yet?

If no, it keeps operating.

If yes, the goal is met and it exits the loop, shutting off the heat.

That is so clear.

So in human psychology, a tote is just the mental feedback loop we run to figure out if our strategy for doing something is actually working.

Yes.

We run these loops constantly without even realizing it.

And what the Strategy Detective app does is deliberately run a user through a tote loop regarding a past success.

Okay, how so?

It guides them back into a specific memory of a time when things went perfectly.

And as it guides them, it helps them elicit their sub modalities.

Whoa, stop right there.

Sub modalities.

That is a very heavy, jargon filled word.

What are we actually talking about there?

Sure.

Think of submodalities as the specific sensory building blocks of a memory.

If you recall a memory right now, your brain doesn't just store, like, raw data.

It stores sensory information.

Like sights and sounds.

Exactly.

So the app asks, when you picture that successful moment, is the image in your mind bright or dim?

Is it moving or still?

Are the sounds loud or muffled?

Oh, I see.

Those sensory details are the submodalities.

The app helps uncover those, along with the hidden beliefs that were driving the success.

Here's where it gets really interesting to me.

The author points out a highly deliberate design choice for this Strategy Detective tool.

She built it with a voice only interface.

Yes.

She entirely avoids text input.

Why does the medium matter so much here?

Why couldn't I just type My answers into a chat window.

Because typing fundamentally alters your brain state, when you were deep inside a memory, visualizing the brightness and hearing the sounds, you were participating in the experience, right?

The moment you open your eyes, look at a keyboard, and try to figure out how to articulate that feeling into a typed sentence, you have pulled yourself completely out of the experience.

Oh, wow.

You shift from being the main character in the memory to being like an outside observer trying to write a report about it.

The analytical part of the brain just takes over.

Exactly.

What's fascinating here is how well Heather understands that friction.

A guiding voice interface allows you to stay inside the state.

You just keep your eyes closed, right?

You listen to the gentle prompts, and you just speak your sensory responses out loud.

The friction of the keyboard is entirely gone.

That is brilliant.

And because you stay in that state, you can successfully build what NLP calls a circle of success.

You identify those positive triggers and create a mental anchor so you can step back into that confidence whenever you need it.

It is an incredibly sophisticated use of technology to facilitate deep psychological work.

But, you know, listening to how capable this AI is, like, it's running the tote loop.

It's dynamically asking about submodalities.

It's guiding the audio.

It leads me to a pretty confrontational.

Question weighed on me.

If the AI is doing all of this heavy lifting, what on earth is the role of the human practitioner?

Like, are the coaches and therapists just making themselves obsolete here?

That is the fear most professionals have, honestly.

But Heather argues the exact opposite.

And this brings us to what she considers the absolute most critical feature of her entire design philosophy.

Okay.

It revolves around her NLP voice model, which basically sits across and oversees all of these tools.

The model has four distinct phases, but the key mechanism is for every single element or pattern the AI proposes.

It offers the human practitioner three explicit.

Keep edit or reject?

Keep edit.

Reject.

And Heather emphasizes in no uncertain terms that the reject button is the single most important part of everything she builds.

Wait, really?

Why is the ability to say no to the machine the most valuable feature?

Because it clearly defines the boundary between machine capability and human intuition.

We have to acknowledge what AI is spectacularly good at.

It is an unmatched engine for rapid pattern generation.

It is very fast.

Extremely.

Given a clean design specification, an AI can produce a hundred variations, distillations or distortions of a concept in less time than it takes you to sip your coffee.

It has limitless stamina for all the tedious work.

But here is its fatal flaw.

It absolutely Cannot calibrate the human being sitting in the room.

If you are working with a client, the AI cannot notice that the client's breathing just became shallow.

It can't see the micro expressions tighten around their eyes.

It cannot sense that the emotional temperature of the room just plummeted.

It's completely blind to that.

Yes, it completely lacks the, the physical and intuitive discernment required for true empathetic human connection.

So the AI is essentially just holding up the scaffolding.

It can formulate the verified patterns and offer you a menu of 50 different directions to take the conversation.

But only you, the human, can step back, look at the client, gauge their breathing, and be the master architect.

Yes, you look at the list and say, no, 49 of those are wrong for this specific person in this specific moment today.

What's fascinating here is that the AI formulates the pattern, but only the human brings the discernment.

If we connect this to the bigger picture, it fundamentally changes where your value lies as a professional.

It's no longer about creation, right?

Your value is no longer in creating the patterns from scratch.

Your value is in your discernment.

The model describes the options, but the master architect decides what to do with them.

The human always, always holds the veto power.

That is a really profound realization.

And honestly realizing that AI is just this lightning fast pattern simulator, desperately in need of human guidance kind of takes the air out of the massive hype bubble surrounding the tech industry right now.

Oh, absolutely.

Like you read the news and it sounds like Silicon Valley is inventing unprecedented magic every single week.

The source text gets quite humorous about this exact dynamic.

Yes, the translations she provides in the text are so validating to read.

It totally strips away the intimidating tech jargon.

For example, tech bros love talking about complex quote unquote file architectures, right?

And Heather just points back to the 70s and says, yeah, we call those logical levels.

It's true.

Across the board, the tech industry boasts about creating autonomous agent loops with built in exit conditions.

And Heather just taps the sign that says tote, test, operate, test, exit.

The exact same feedback loop just applied to code instead of a brain.

Exactly.

Or take the concept of a test suite in software engineering, which is a systematic way to test boundaries and identify missing data.

NLP has been teaching that exact methodology for decades under the name the Meta model.

Or when software engineers talk about conducting a requirements interview to figure out what A system needs to achieve an outcome.

In human psychology, that is literally just basic strategy elicitation.

You're interviewing the system to find the requirements.

It's so funny if we connect this to the bigger picture, the tech industry is furiously patting itself on the back, convinced they have birthed these entirely new scientific disciplines called context engineering or agent evaluation.

Right.

Meanwhile, the psychologists are sitting there saying, we've been doing this for 50 years.

You just gave it a digital interface.

It completely demystifies the technology.

It proves that the human frameworks we already naturally possess are actually the perfect blueprints for managing and designing AI.

So what does this all mean for you, the listener?

How do we take this massive paradigm shift and actually apply it?

Well, the text highlights a very tangible opportunity, specifically for educational programs or training schools.

Because you now know how to map your existing curriculum onto these digital frameworks, you can confidently use AI to handle all the tedious practice and repetition that happens between live training modules.

Right.

So you deploy that 11pm language game we talked about to handle the homework.

Yes.

And because it's running off your customized AI brain, it maintains your curriculum, your unique voice, and your specific rules.

But, and this is the critical distinction here, you keep the actual core learning experience and that irreplaceable human calibration strictly inside the live training event.

You outsource the repetition, but you fiercely protect the calibration.

That is such a brilliant way to structure a business.

This whole deep dive really has been quite a journey.

It really has.

We started with the frustrating reality of passively feeding prompts into a generic black box, just hoping for a text output that didn't sound like a robot.

And we've arrived at the ability to deliberately architect an ecosystem that preserves your identity, your sub modalities, and all your messy human quirks.

Totally immune to whatever AI platform happens to be trending on Twitter this week.

Exactly.

It is fundamentally about taking ownership.

It is.

And to that end, before we finish up today, I want to pass along a direct challenge that Heather poses to her readers at the end of part one.

Think of this as a small task for you to complete today.

Oh, I'm always down for a practical application.

What is the task?

I want you to think about one specific mental framework or a process you use in your daily work that you would entirely trust an AI to recreate and handle for you.

Okay.

Identify one thing you'd confidently hand over,.

Then think of one framework, process or interaction that you would never hand over to an AI.

Oh, I like that.

But don't stop at just naming it, you need to articulate exactly why that never task matters too much to outsource.

As the author notes, that second answer, the deep why behind the thing you refuse to let go of, might just be the most important insight you have all year.

Because that why essentially defines your ultimate value as a human being in your profession.

I absolutely love that.

So, as you go about your day thinking about what you are willing to hand over and what you will stubbornly guard with your life, I want to leave you with a final thought to mull over.

Go for it.

If we are entering an era where we can perfectly encode our quirks, our core values, our idiosyncratic phrases, and our negative constraints into a swappable digital environment, yeah.

Does that make our digital clones more human?

Or does the sheer perfection of that mimicry force us to completely redefine what makes us irreplaceable in the first place?

If a machine can flawlessly generate your messy human stock photo, what does your real uncodable X ray look like now?

Wow.

That is a question we are all going to have to answer.

Likely much sooner than we think.

Truly.

Thank you for joining us on this deep dive.

We'll see you next time.